Comment 8 for bug 183652

You were right Till, the bug is related not related to any of the recent updates of cups but to the fact that cups is starting to early for avahi to be ready.

The bug is present on only one of my numerous machines (many others have no symptoms so this really seems schedule-related)
On this machine I have reproduced it with all 1.3.2-1ubuntu7* cups versions (from 1.3.2-1ubuntu7 to 1.3.2-1ubuntu7.5), so it is confirmed the regression came with avahi.
The bug disappears either by reversing the effect of the avahi 0.6.20-2ubuntu3.1 update ( with : for l in 2 3 4 5; do old=/etc/rc$l.d/S18avahi-daemon; new=/etc/rc$l.d/S24avahi-daemon; if [ -e $old ] && ! [ -e $new ]; then sudo mv $old $new; fi; done )
Or by introducing a sleep time in the init file of cups.

Before the fix on the avahi side of bug 173470, avahi was starting too late, so cups was working but was opting at each start to not use avahi at all. Now it is using avahi, but it may not start depending of the timing between avahi and cups.

The bug should be either solved in the avahi startup script as Till said, or by introducing in the scheduler of CUPS a regular poll of the avahi-daemon (similar to the poll for new network interfaces that 1.3.2-1ubuntu7.5 has made come back) that would allow CUPS to dynamically change his mode of working (with or without avahi)